I find the InfoGain loss layer that is now available in caffe (https://github.com/BVLC/caffe/pull/3855 ) very useful for my multi-class classification application with highly imbalanced data distribution. It would be wonderful to have InfoGain loss layer in the nvcaffe so I can get the best out of the GPUs and DIGITS. Can someone help me in implementing that in nvcaffe or a new feature request for nvcaffe. Thanks in advance for the help.
I find the InfoGain loss layer that is now available in caffe (https://github.com/BVLC/caffe/pull/3855 ) very useful for my multi-class classification application with highly imbalanced data distribution. It would be wonderful to have InfoGain loss layer in the nvcaffe so I can get the best out of the GPUs and DIGITS. Can someone help me in implementing that in nvcaffe or a new feature request for nvcaffe. Thanks in advance for the help.